Modeling Physiological and Behavioral Data Streams Towards Health Insights

Advancements in wearable computing have minimized the level of invasion necessary for digital sensors to passively collect aspects of one's daily life. The fine grained physiological and behavioral measurements these devices provide serve as invaluable tools for the fields of health and wellness. Furthermore, these devices provide a novel and personal manner for individuals to interact with their health data. With the field of wearable technology still largely in its infancy, this relationship is limited to these devices serving only as digital mirrors of one's behaviors: providing daily step counts and sleep duration, yet leaving the user to determine what these numbers mean for their health. It is here that this dissertation aims to advance this relationship by utilizing the data streams generated by these devices to infer users health through their health behaviors. This is accomplished by transforming these streams to extend the range of behaviors they capture, measuring changes in behaviors and physiology overtime, and addressing the challenges of use and abandonment of these devices. By successfully informing users of their health through their health behaviors, wearable devices will serve as a pervasive and convenient platform for engaging with personal health.